Your AI tool just turned a two-hour task into 20 minutes. Congratulations. Now, what happened to the other hour and 40 minutes?
New research from Wharton suggests many organizations aren’t asking. They’re counting the hours saved instead of what those hours produce.
Every technological transformation has promised to make us faster, more productive, and more efficient. Many have delivered (sometimes in ways that nobody originally envisioned).
Over four decades in technology, I’ve learned that just because technology makes a task easier doesn’t mean it makes the outcome better. That’s why every business leader, risk manager, and workers’ compensation professional should pay attention to Wharton’s latest study.
Wharton recently released its fourth annual study of enterprise AI adoption, The Adaptive Advantage: Who’s Winning with AI and How, conducted with GBK Collective. Researchers surveyed 1,416 people at large U.S. companies, from C-suite executives to entry-level employees. I’ll share my takeaways below, but I’d encourage every leadership team to read the full report.
The adoption numbers are impressive.
- 62% of managers report using AI daily
- 91% use AI more than weekly
- 73% say AI has significantly changed their workflows
We’re well beyond the days when AI was something interesting to experiment with during lunch. It’s becoming part of how organizations operate.
While 80% report positive returns from AI investments, only 28% report strongly positive returns. And just 7% of organizations demonstrate success across all three measures Wharton identified: proving value, transforming workflows, and aligning the organization.
Organizations are adopting AI at extraordinary speed. They’re spending money, changing processes, and introducing new tools. But relatively few are putting all the pieces together. Using AI and succeeding with AI are two very different things.
One statistic particularly deserves attention.
Forty percent of departments measure AI’s return on investment by hours saved, yet only 27% use net-new revenue as a measure.
Saving time is certainly valuable. I use AI daily for both professional and personal needs. From developing content to crafting meeting takeaways to simplifying research, I can attest to its power. And more than once, I’ve caught it being confidently wrong.
So, back to that hour and 40 minutes. Did the quality improve? Did the customer receive better service? Did the employee use that additional time to solve a more important problem? Or did we simply get the same result faster?
Wharton found that 35% of departments measuring at least one growth-related outcome reported strongly positive returns, compared with just 13% of those measuring efficiency alone.
Whether correlation or causation, that difference is hard to ignore.
There’s a difference between improving productivity and creating value. I suspect many organizations haven’t fully considered it.
Imagine a claims adjuster using AI to summarize medical records. Instead of spending an hour reviewing documents, the adjuster receives a summary in minutes. That’s a tangible efficiency improvement.
But did the summary accurately identify the important medical issues? Did it recognize conflicting information? Did the adjuster verify the output? Did that saved time translate into better communication with the injured worker, a more thoughtful claim strategy, or a faster return to work?
Now consider a risk manager using AI to identify emerging workplace safety trends. If technology helps recognize a pattern of injuries before another employee gets hurt, that’s not efficiency. That’s an employee who didn’t get hurt.
The same applies to identifying claims that are beginning to deteriorate, improving employee training, or helping smaller organizations access capabilities they previously couldn’t afford. The real opportunity isn’t simply automation, but using technology to improve decisions, prevent problems, and produce better human outcomes.
Then there’s the finding on agents.
Sixty-seven percent of departments report using AI agents in some form. These are systems that can go beyond answering questions to perform tasks or take actions, such as sending a denial letter or setting a reserve. Only 17% report allowing agents to operate without someone approving every step.
We’re moving from AI that recommends to AI that acts.
What happens when an AI system influences a medical treatment recommendation, a claim reserve, or a decision about an employee’s benefits? Who …
- verifies the information?
- recognizes an error?
- is accountable when something goes wrong?
Oh, and who keeps track of your vendors’ use of AI?
I’m not convinced most organizations are developing the knowledge, policies, oversight, and accountability necessary to use AI responsibly.
Wharton’s research reinforces that concern. The report found that organizational alignment requires at least four of five factors: adequate AI training, a clear long-term vision from leadership, meaningful employee choice over how and when to use AI, people and processes that don’t hold the department back, and little resistance among colleagues. Notice that none of them is about the software.
And there’s a telling gap on the first one: 78% of leaders believe employees receive adequate AI training. Only 56% of employees agree.
Successful implementation involves more than purchasing software. It requires changing workflows, educating employees, establishing expectations, and aligning leadership around a common purpose. That’s a management challenge, not simply an IT project.
Next week, I’ll be in Sacramento delivering the keynote at the California Alliance of Self-Insured Groups’ annual conference. My presentation is titled “AI Is Already in Your Organization. The Only Question Is Who’s Controlling It.”
I chose that title deliberately.
Whether you’re running a self-insured group, managing claims, operating a business, or providing services to employers, AI is probably influencing your operations already.
Every leader should be able to answer four questions: Where is AI being used? What is it allowed to do? How are we measuring its value? Who’s responsible for the results? In reality, every employee should be asking them too.
I’m an optimistic realist. There are certainly challenges and unknowns as AI becomes more powerful at an ever-increasing pace. However, I believe AI presents extraordinary opportunities.
But enthusiasm without discipline isn’t a strategy. Neither is measuring success exclusively by how many hours we’ve saved.
Here’s where I’ve landed: our goal for AI shouldn’t be to make people unnecessary. It should be to make people more capable.
So the next time someone tells you how much time AI is saving your organization, ask a follow-up question: What are we doing with that time, and how do we know the results are better?
Hours saved is a number. An injured worker who returns to work sooner is an outcome. Start measuring the second one.
